A critical review on computational simulation and machine learning in phonon transport

Abstract

For the last few decades, the study of heat transport properties of different materials has become prominent in a variety of applications ranging from thermoelectric energy conversion to thermal management of electronics. This critical review highlights the advancement of different computational techniques to investigate heat transfer focusing mostly on phonon transport in semiconductors and insulators. These techniques provide an insight into phonon mediated thermal transport, and each technique has its own strengths and limitations in terms of efficiency, accuracy, and scalability. We explore how different methods perform in the calculations of phonon transport properties when comparison is made with the experimental results. This review also presents the integration of the state-of-the-art big data or data driven technology, namely artificial intelligence/machine learning (ML) approach, with different algorithms which has great potential to revolutionize phonon transport research with fractional computational cost, providing fast yet highly accurate predictions of comprehensive phonon properties. The ML approach is also expected to transform the high-throughput study regime by screening and pinpointing materials with desired phonon properties from vast material space. Some key research efforts that have successfully utilized this technique are comprehensively reviewed and future trends and challenges are also discussed.